| In recent years,the balance of non-performing loans in China has been on the rise,while credit business has been continuously expanding,posing significant challenges to the field of financial risk control.Predicting customer default tendencies is particularly important in reducing the economic losses incurred by banks and other financial institutions due to non-performing loans.Currently,ensemble learning is gradually being widely used in the field of financial forecasting,which can leverage the advantages of multiple base learners to improve the prediction accuracy and robustness of the model.Therefore,this study researches and implements a loan default prediction method based on ensemble learning,which is applied to practical business by designing and implementing an automated credit system.Firstly,this thesis has been proposed an improved TabNet model for loan default prediction.Based on the preprocessing of user loan datasets,the shortcomings of the neural network model TabNet in processing user loan data are explained.The model structure is improved by adding multihead attention mechanisms,which enables better handling of highdimensional and dense datasets.At the same time,the convergence speed of the model is accelerated through self-supervised pre-training.Finally,the performance of the improved model is tested through comparative experiments,and the AUC value is increased by 3%.Secondly,this thesis has been proposed a loan default prediction method based on ensemble learning.Based on the Stacking ensemble strategy,the XGBoost model and the improved TabNet model are used as base learners,and LR is used as the second-layer learner to construct a loan default prediction ensemble model.Finally,the performance of the ensemble model is verified through comparative experiments,and all indicators are significantly improved compared to a single model.Finally,based on the above work,an automated credit system was designed and implemented.The core functional modules of the system include loan application,credit approval,credit model management,and credit product management.The system can effectively help banks and financial institutions manage customer loan information and evaluate borrowers’ default tendencies to reduce losses caused by bad loans. |
